Gene Therapy (2004) 11, 856–864
& 2004 Nature Publishing Group All rights reserved 0969-7128/04 $25.00
www.nature.com/gt
RESEARCH ARTICLE
Silencing of episomal transgene expression
by plasmid bacterial DNA elements in vivo
ZY Chen, CY He, L Meuse and MA Kay
Departments of Pediatrics and Genetics, Stanford University School of Medicine, Stanford, CA, USA
We previously demonstrated that sustainable enhanced
levels of transgene products could be expressed from a
bacterial DNA-free expression cassette either formed from
a fragmented plasmid in mouse liver or delivered as a
minicircle vector. This suggested that bacterial DNA
sequences played a role in episomal transgene silencing.
To further understand the silencing mechanism, we systematically altered the DNA components in both the expression
cassette and the bacterial backbone, and compared the gene
expression profiles from mice receiving different DNA forms.
In nine vectors tested, animals that received the purified
expression cassette alone always expressed persistently
higher levels of transgene compared to 2fDNA groups. In
contrast, animals that received linearized DNA by a single
cut in the bacterial backbone had similar expression profiles
to that of intact plasmid groups. All three linear DNAs formed
large concatemers and small circles in mouse liver, while
ccDNA remained intact. In all groups, the relative amount of
vector DNA in liver remained similar. Together, these results
further established that the DNA silencing effect was
mediated by a covalent linkage of the expression cassette
and the bacteria DNA elements.
Gene Therapy (2004) 11, 856–864. doi:10.1038/sj.gt.3302231
Published online 18 March 2004
Keywords: linearized DNA; expression cassette; human a-1 antitrypsin; human factor IX; bacterial backbone; transcriptional
silencing
Introduction
The short duration of transgene expression is an
important obstacle to overcome before nonviral vectors
become useful for a wide variety of gene therapy
applications. Some progress has been made in understanding the mechanism underlining the loss of transgene
expression, as well as in developing the methodology to
prolong the production of the therapeutic gene product. It
has been found that after transfection of liver, muscle and
lung in vivo, the transgene product from the plasmid
DNA is typically expressed for a short period of time,
even though vector DNA is not lost.1–6 These observations
are consistent with the hypothesis that gene silencing,
rather than loss of vector DNA, is the primary mechanism
limiting long-term episomal transgene expression.2,6,7 The
immunogenic CpG dinucleotides in the bacterial backbone of plasmid, and the interaction between the vector
DNA elements and a variety of cytokines have been
suggested as playing key roles in episomal gene silencing.6–10 At the same time, some prolongation of
transgene expression from nonviral vectors has been
achieved. For example, by using a composite promoter/
enhancer in the context of a minigene, therapeutic levels
of human factor IX (human FIX) from mouse liver were
achieved for up to 18 months.11 A high level of transgene
expression was maintained for at least 8 weeks in the
Correspondence: MA Kay, Departments of Pediatrics and Genetics, 300
Pasteur Dr. Rm G305, Stanford, CA 94305-5208, USA
Received 24 June 2003; accepted 20 November 2003; published
online 18 March 2004
lung after delivery of a reporter gene attached to the
human polyubiquitin C promoter.2 In another case, Yew et
al12 used a hybrid promoter composed of cytomegalovirus (CMV) promoter and ubiquitin B promoter combined with the deletion of CpG dinucleotides in
kanamycin resistance gene, and shortening of plasmid
origin of replication in bacterial backbone, to express high
and sustainable levels of human alpha galactosidase A
from murine liver for up to 35 days. Using CMV or
chimeric promoter, Liu et al13 and Kramer et al14 also
achieved prolonged and high levels of transgene products from mouse liver. Nevertheless, the mechanism of
the promoter inactivation remains poorly defined, and
the development of vectors capable of expressing high
level of transgene products so far remains as the result of
occasional findings. A thorough understanding of the
mechanism underlying the episomal vector silencing
effect will be important for the successful development
of versatile nonviral vectors that can be used to achieve
persistent gene expression in different cell types.
Recently, we demonstrated that the transcriptional
silencing phenomenon observed with plasmid delivery
in vivo could be decreased or abolished by cleaving a
closed circular plasmid into two DNA fragments before
delivery into mice.1 Persistent high levels of serum
human a1-antitrypsin (human AAT) or human FIX, were
obtained by cutting the expression plasmid to separate
the expression cassette from the bacterial backbone prior
to transfection of mouse liver, while the equivalent molar
amount of ccDNA resulted in 10–100 times less gene
expression. Analysis of the molecular structure of the
injected linear vector DNA in mouse liver demonstrated
Bacterial DNA in episomal transgene silencing
ZY Chen et al
that the total amount of vector DNA was similar in both
groups of mice, but linear DNA resulted in the formation
of large random concatemers as well as smaller circles,
while the ccDNA remained as intact circular structures.
Accordingly, we have considered the possibility that the
large concatemers might be responsible for the enhanced
transgene expression, or the increase in transgene
expression might be a consequence of the distance or
dissociation of the expression cassette from the bacterial
DNA, which might influence its inhibitory role in
transgene expression.1 Here, we present additional
evidence that the bacterial backbone is involved in
silencing the transgene of episomal vectors in vivo. We
also demonstrate that a high and sustainable level of
transgene expression can be achieved by simply excluding bacterial DNA and using a purified expression
cassette. Taken together, our observations have laid
down a foundation for further understanding the
molecular mechanism responsible for the inhibitory
effect of the bacterial DNA, and the development of
nonviral vectors capable of expressing sustainable high
level of transgene and suitable for use in human gene
therapy.
Results
High level of transgene expression by infusion
of the purified expression cassette
To begin to unravel the mechanisms whereby there is a
discordant level of gene expression from the same DNA
sequences delivered as linear fragments versus circular
DNA molecules, we investigated two different possibilities: (1) the bacterial backbone played an inhibitory role
in transgene expression, which was lost or weakened
when separated from the expression cassette during
concatemer or small circle formation, and/or (2) the
structure of large concatemers was more favorable for
transgene expression. We compared transgene expression from mice receiving a purified expression cassette to
those receiving two-linear fragments consisting of the
expression cassette and bacterial plasmid backbone. We
reasoned that if bacterial DNA plays an inhibitory role in
transgene expression, mice transfected with the purified
expression cassette free of bacterial DNA would express
higher levels of the transgene than mice receiving the
expression cassette together with bacterial DNA. We
infused mice with 20 mg of uncut circular plasmid
pRSV.hAAT.bpA (ccpRHB), two-fragment linear DNA
(2fRHB) or different doses of purified expression cassette
(1fRHB), and compared serum human AAT levels at
various time points (Figure 2a). Consistent with previous
observations,1 the uncut circular plasmid injected mice
expressed a high level of serum human AAT that
declined by more than 3 logs within 4 weeks after
DNA infusion. In contrast, mice receiving equal molar
amounts of 2f- or 1fRHB had 28- and 40-fold higher
serum human AAT, respectively, over the same period.
The serum human AAT levels were approximately
proportional to the vector DNA doses in the 1fRHB
groups. In all, 20 mg of 2fRHB contained 8 mg of the
RSV.hAAT.bpA expression cassette. However, serum
human AAT levels from the mice receiving the same
molar amount of the expression cassette delivered as
2fRHB were 40–60% lower than that of mice receiving the
expression cassette alone. These data suggested that a
substantial proportion of the human AAT expression
cassette was silenced by the bacterial backbone and
support the hypothesis that bacterial DNA sequence has
a negative effect on transgene expression.
857
Covalent linkage of bacterial DNA is responsible
for transgene silencing
Infusion of the 2fRHB DNA vectors into mouse liver
resulted in the formation of random concatemers.1 To
differentiate whether the concatemers or the covalent
linkage of the bacterial and expression cassette sequences
were responsible for the fall off in gene expression, we
compared serum human AAT levels from XbaI-linearized DNA (LRHB) and 2fRHB DNA injected mice. XbaI
cuts once through the bacterial sequence (Figure 1),
resulting in a linear expression cassette covalently
attached to the bacterial DNA. We infused each mouse
with 40 mg LRHB, 2fRHB, uncut ccpRHB or an equal
molar amount of 1fRHB. Interestingly, the serum human
AAT level in the LRHB group was similar to that of the
mice receiving the ccpRHB (Figure 2b). As demonstrated
in the above experiments, mice receiving the 1fRHB
expressed 40–150% more transgene than the 2fRHB
group, which was 20- to 55-fold higher than the LRHB
or ccpRHB groups, 3 weeks after DNA infusion. These
observations further confirm that the bacterial DNA
had an inhibitory effect on transgene expression, and a
covalent attachment between these two DNA elements
was required for the silencing effect to occur.
It was possible that the fall off in transgene expression
in the mice receiving LRHB mimicked that of ccpRHB
because the DNA recircularized and formed ccpRHBlike molecules in mouse liver. To answer this question,
we set forth to determine the molecular structure of the
vector DNA in mouse livers by Southern blot analysis.
Mouse liver DNA was digested with either BglII, which
did not cut the vector DNA but cut the mouse genome
frequently, or HindIII, which cut vector DNA once
through the expression cassette (Figure 1). In the blot of
liver digested with the 0-cutter BglII and probed with
radio-labeled human AAT cDNA, strong vector DNA
bands over 23 kb in size were found in all of the mice
receiving 1fRHB, LRHB or 2fRHB linear DNA (Figure
3a). This established that all three linear DNA species
formed large concatemers in mouse livers. Multiple
DNA bands of smaller sizes were also seen in all three
linear DNA groups, indicating that DNA circles with
different numbers of DNA fragments were formed. A
1.6 kb band, seen in the 0-cutter blot, was replaced by a
2.1 kb band in the one-cutter blot in 1fRHB and 2fRHB
groups, but was absent in the other two groups,
indicating that this 1.6 kb band represented the recircularized expression cassette monomer. Consistent with
previous observations,1 the one-cutter HindIII converted
all of the vector DNA bands observed in the 0-cutter BglII
blot of the 2fRHB group into a DNA ladder, ranging from
2.1 kb to up to 23 kb (Figure 3a). Apparently, each band
was composed of one or two hAAT.bpA sequences, with
variable numbers of bacterial DNA sequences, suggesting that the large concatemers were formed by random
linking of the two DNA elements in mouse liver. In
contrast, only two DNA bands were found in the LRHB
and 1fRHB groups. The two vector DNA bands, about
Gene Therapy
Bacterial DNA in episomal transgene silencing
ZY Chen et al
858
Figure 1 Schematic illustration of nine DNA vectors expressing either human AAT or human FIX. Amp, ampicillin-resistant gene; CM, chloramphenicolresistant gene; ColE1, 15A and MB1, plasmid origins of replication; F1( þ ), f1 phage origin of replication; abbreviation of restriction enzymes: Bs, BspH1;
E, EcoR1; H, HindIII; S, SpeI; X, XhoI; Xb, XbaI. The DNA elements different from that in pRSV.hAAT.bpA are highlighted by using the filled symbols.
Figure 2 Serum human AAT levels from mice injected with different doses of purified RSV.hAAT.bpA expression cassette (1fRHB), 2fRHB, LRHB or
uncut closed circular DNA (ccpRHB). (a) Each mouse received 32 ng to 32 mg of 1fRHB, or 20 mg of 2fRHB or ccpRHB, which represents the same amount
of expression cassette equal to 8 mg of 1fRHB (n ¼ 5 mice per group). (b) Each mouse received 40 mg of 2fRHB, LRHB, ccpRHB or an equal molar amount of
1fRHB (16 mg of the pure RSV.hAAT.BpA fragment per mouse, n ¼ 5 mice per group).
3.4 and 5.0 kb in the LRHB samples, and 2.1 and 3.4 kb
in the 1fRHB group, represent the HindIII cleavage
products of head to tail, and tail to tail sequences in each
group. When the blot was probed with a full-length
Gene Therapy
expression cassette instead of human AAT cDNA, a third
band of about 0.8 kb, and 6.4 kb in the 1fRHB and LRHB
groups, respectively, was found (data not shown),
indicating the existence of head to head junctions. These
Bacterial DNA in episomal transgene silencing
ZY Chen et al
859
Figure 3 Southern blot analysis of vector structure in mouse livers. (a) In all, 20 mg of liver DNA from mice receiving the four forms of DNA, as indicated
in the legend of Figure 2b and killed 11 weeks post-DNA injection, was digested with either the 0-cutter Bgl II (left panel) or the one-cutter HindIII (right
panel). The blot was probed with a radio labeled 1.35 kb EcoR1 fragment of hAAT cDNA as indicated in Figure 1. (b) The liver DNA was digested with
BglII in the upper panel and with HindIII in the lower panel. The liver DNAs were from the same group of mice as indicated in the legend of Figure 2a and
killed at week 11. All the experimental conditions were the same as described in (a). (c) Phosphoimager quantification of vector DNA in liver DNA samples
used in (a). *Indicates the difference in the copy number per diploid genome is statistically significant as compared to 2fRHB or LRHB group at Po0.01,
and o0.01.
observations strongly suggest that linear DNAs link
together randomly, probably via a nonhomology end
joining mechanism,15 to form large concatemers or
smaller circles. As the LRHB formed large concatemers
and expressed a 20–55 times lower level of serum human
AAT, with a pattern closely similar to that of the ccpRHB
group (Figure 2b), our observations strongly suggest that
formation of large concatemers was not responsible for
the high levels of transgene expression in 1fRHB- and
2fRHB-treated animals.
When the structure of vector DNA in the livers of mice
receiving different amounts of 1fRHB were analyzed by
Southern blot, a very strong vector DNA signal 423 kb
was detected in mice receiving 32 mg of 1fRHB, while a
much weaker signal was detected in 8 mg 1fRHB-treated
animals. No signal was detected in mice receiving lower
doses of linear DNA (Figure 3b). In the mice receiving
less than 0.5 mg each of 1fRHB DNA, a single band was
observed in BglII- (no cutter) or HindIII- (single cutter)
digested DNA, indicating that the vector formed a
circularized RSV.hAAT.bpA expression cassette monomer. These observations suggested that formation of
large concatemers occurred only when the linear DNA
concentration reached a threshold in the hepatocytes.
These results are not unexpected, because at low DNA
concentrations intramolecular ligations are likely favored
over intermolecular ligations. Since low but stable levels
of serum human AAT were detected in all mice receiving
three low doses, 32–500 ng each, of 1fRHB (Figure 2a),
this suggested that the circular expression cassette
monomer was the transcriptionally active DNA form
in vivo. This view was furthermore confirmed by the
observations that minicircle vectors could express robust
levels of transgene reporters persistently.16
To determine if a difference in the amount of vector
DNA in mouse liver contributed to the variation in
transgene expression between the different experimental
groups, liver DNA from mice receiving four different
forms of DNA, as indicated in the legends of Figures 2b
and 3a, were used for the determination of vector DNA
Gene Therapy
Bacterial DNA in episomal transgene silencing
ZY Chen et al
860
copy number (Figure 3c). Consistent with our previous
observation,1 about four to seven copies of vector DNA
per diploid genome were detected among the 2fRHB,
LRHB and ccpRHB groups. However, significantly less
vector was observed in 1fRHB compared to 2fRHB- and
LHRB-treated mice (Po0.01). Since higher amounts of
transgene product were expressed from significantly less
1fRHB DNA in mouse liver, these results exclude the
possibility that differences in transgene expression
resulted from the difference in the amount of vector
DNA in mouse liver. Taken together, these results
suggest that it is the distance or dissociation of the
expression cassette from the bacterial backbone that
results in high levels of transgene expression from
RSV.hAAT.bpA, and the formation of large concatemers,
per se, did not contribute to the maintenance of high
levels of transgene expression.
ampicillin resistance gene and the pMB1 DNA origin of
replication, which was highly homologous to ColE1 of
pBS.KS. To determine if any of the bacterial sequences in
common between the plasmids were responsible for the
transcriptional inhibition, the DNA origin ColE1 was
replaced with p15A from pACYC184, and the ampicillin
resistance gene was replaced by the chloramphenicol
(CM) resistance gene from pBC.KS. Transgene expression
pattern from uncut plasmid DNA of these two constructs
was compared with that from mice receiving either
2fRHB or uncut ccpRHB (Figure 4c). Both uncut circular
pCM.RHB and p15A.RHB demonstrated a serum human
AAT expression pattern similar to that of ccpRHB. These
two experiments suggest that the negative effect of
bacterial backbone was independent of a specific bacterial
DNA sequence.
Bacterial DNA inhibitory effect was promoter/enhancer
independent
It is possible that the bacterial DNA sequences present in
plasmid DNA do not universally inhibit transgene
expression from all expression cassettes. To determine
if other sequences might be refractory, we used four
other promoter/enhancers, including two viral promoters, CMV and SV40, and two cellular promoters,
heI4AF1 and glyceraldehyde 3-phosphate dehydrogenase promoter (GAPDH), in place of the RSV promoter.
Purified expression cassette, two DNA fragments and
uncut ccDNA were infused into groups of mice and
serum human AAT levels were compared (Figure 4a).
Additional promoters with the human FIX expression
cassettes in vectors psApoE.hFIX þ Int A.bpA and
pEF1a.hFIX.hGHpA were also tested (Figure 4b). As
expected, serum transgene protein levels were different
with the various promoters. The serum human AAT
levels, from the respective purified hAAT expression
cassettes varied by up to more than 100-fold. Furthermore, uncut plasmids psApoE.hFIX þ Int A.bpA,
pGAPDH.hAAT.bpA and pCMV.hAAT.bpA expressed a
low level of serum human FIX or human AAT throughout the 11- or 20-week time period, while the transgene
expression from uncut pEF1a.hFIX.hGHpA and
pheIF.hAAT.bpA plasmids declined to undetectable
levels within 7 weeks. While some of these observations
may reflect the differences in transgene expression
activity from different promoter/enhancers, a general
expression pattern was revealed from all six constructs.
All six purified expression cassettes expressed a higher
level of transgene product than their two DNA fragment
counterparts, and all of the uncut circular plasmids
expressed the lowest levels of transgene product shortly
after DNA infusion. The serum transgene reporter from
mice receiving one fragment DNA could be 2- to 3-log
higher than that of the uncut plasmid DNA group
(Figure 4a). Thus, our data demonstrate that the
inhibitory effect of bacterial DNA sequence was reporter
and promoter/enhancer independent.
Discussion
Different bacterial DNA sequences had the same
inhibitory effect
Five of the six constructs tested were based on the
plasmid backbone pBS.KS, while the sixth construct,
although containing different sequences, had the same
Gene Therapy
In this study, we tested a total of nine constructs with
seven different expression cassettes. All of the data
indicate that covalent linkage of the bacterial DNA
silenced the transgene expression cassette in episomal
vectors in mouse livers. Consistent with our previous
observation that minicircle DNA vectors expressed
sustainable high levels of transgene products,16 exclusion
of bacterial DNA by using purified expression cassette
resulted in up to 2- to 3-log higher levels of transgene
expression as compared to their standard plasmid
counterparts in mouse livers. In agreement with previous observations,1 the enhanced transgene expression
also occurred in two fragment (linear expression cassettes and bacterial plasmid backbone) DNA groups, as
compared to intact circular plasmid groups.
Importantly, the difference in transgene expression
activity was not due to the difference in vector DNA in
mouse livers because, as we have demonstrated previously,1,16 the absolute amount of vector DNA was
similar regardless of which form of the vector was
infused. Although linear DNA molecules can concatemerize in vivo, our data indicate that concatemerization
of vector DNA, per se, did not support transgene
expression as we speculated earlier.1 Linearized DNA
generated by one cut through the bacterial DNA also
formed large concatemers and demonstrated a similar
depressed transgene expression profile as the uncut
plasmids. Our data also suggest that concatemerization
was not necessary for sustained high levels of transgene
expression. In the two fragment DNA groups, expression
cassette circles free of bacterial DNAs were formed
alongside large concatemers. In the light of our earlier
observations of the robust transgene expression profiles
of minicircle vectors,16 it appears that it was the bacterial
DNA-free recircularized expression cassettes alone or in
combination with large concatemers with one or more
expression cassette units separated from the bacterial
backbone sequence that remained transcriptionally active. Taken together, our data further suggest that a
covalent connection between the bacterial DNA and
expression cassette was necessary for transgene silencing
in vivo.
The molecular mechanism underlining the bacterial
DNA silencing is not known at present. However, based
on the data of the present study, as well as others, several
Bacterial DNA in episomal transgene silencing
ZY Chen et al
861
Figure 4 Serum human AAT or human FIX from mice receiving one of the three forms of vectors (1fDNA, 2f DNA or ccDNA) with different DNA
components. (a) Serum human AAT from mice receiving 40 mg of 2fDNA, ccDNA or an equal molar amount of purified expression cassette (1fDNA) from
pGAPDH-, pheIF-, pCMV- or pSV.hAAT.bpA. (b) Serum human FIX levels from mice receiving 20 mg of two-fragment DNA (2fDNA), uncut ccDNA or
an equal molar amount of purified expression cassette (1fDNA), from either p.EF1a.hFIX.bpA (EF1a) or psApoE.HCR.hAATp.hFIX þ Int A.bpA (sApoE).
(c) Serum human AAT from mice receiving 20 mg of 2fRHB, or ccpRHB derived from vectors containing the p15A (replacing the ColE1 DNA), or CM
resistance gene (replacing the ampicillin resistance gene) (n ¼ 5 mice per group).
Gene Therapy
Bacterial DNA in episomal transgene silencing
ZY Chen et al
862
possibilities are worthy of serious consideration. One
prominent functional feature common to all the bacterial
backbones, for example, is that they are transcriptionally
inactive in mammalian cells, although the eukaryotic
expression cassettes in cis- are active. Functionally, these
transcriptionally inert DNA sequences are similar to the
heterochromatin in eukaryotic cells and we speculate
that they may also share a similar chromatin structure.17
Differences in many molecules composing the heterochromatin and euchromatin have been defined. For
example, nonacetylated histones 3 and 4 bind DNA
tightly and are the features of heterochromatin, while
acetylation of these two histones will loosen their
binding to DNA, allowing the formation of more open
euchromatin.17 The machinery responsible for the maintenance of chromatin status has also been studied in
depth. It has been demonstrated that components of the
transcription complex have acetylase activities, which
are responsible for the addition of acetyl group to
histones 3 and 4.18 Since the bacterial DNA is not
transcribed and lacks such a mechanism to maintain the
acetylation status of histones 3 and 4, the chromatin in
this region may be more condensed. If this is the case, it
may be possible to unravel the mechanism of the
inhibitory effect exhibited by the bacterial sequences.
It has been well documented that there are insulators
between euchromatin and heterochromatin, and deletion
of these insulators will result in the spreading of the
heterochromatin into the neighboring euchromatin, and
consequently the silencing of the genes.19 As no known
insulator was included in all the constructs tested in this
study, heterochromatin of the bacterial backbone, if
formed, could spread freely into the expression cassette,
resulting in the silencing of the uncut circular plasmid. In
contrast, in a scenario such as when the bacterial DNA
and eukaryotic expression cassette were transfected as
two molecules, heterochromatin spreading and transgene silencing would be limited to those molecules that
remained physically connected with the bacterial sequence. Thus, a substantial number of the expression
cassettes, for example, those that formed small circles
free of bacterial DNA, and/or were distanced from
bacterial DNA in large concatemers, could maintain a
more open euchromatin status and a higher rate of
transcription activity. Kass et al20 have demonstrated the
spreading of a ‘repressive nucleoprotein structure’ in the
center of a methylation site into a neighboring promoter,
resulting in the silencing of a plasmid injected into
Xenopus oocytes.
The influence of cytosine methylation status in the
vector DNA on the transgene expression is worthy of
careful consideration. The frequency of CpG dinucleotides is much higher in bacterial DNA than in vertebrate
DNA.10 Consistent with this hypothesis, in the plasmid
pRSV.hAAT.bpA used in this study, the frequency of
CpG dinucleotides in the ampicillin resistance geneColE1 origin region (6.4 CpG per 100 bases) is 2.6 times
of that in the RSV.hAAT.bpA expression cassette (2.8
CpG per 100 bases). A similar higher frequency of CpG
was also seen in the CM resistance gene and the p15A
DNA origin (4.9, and 5.8 per 100 bases, respectively). In
this study, the CpG dinucleotides in plasmid DNA
derived from the bacterial strain (XL1-blue F0 of
Stratagene) were unmethylated; however, we cannot
rule out methylation, which might occur after the DNA
Gene Therapy
was introduced into the liver. Nonetheless, we demonstrated that the inhibitory effect of the bacterial backbone
was DNA sequence independent. All three of the
bacterial backbone elements, including the F1 phage
origin of replication, plasmid origin of replication and
antibiotic resistant gene, were systemically tested by
either being deleted or placed in different combinations.
All four plasmids demonstrated the same depressed
transgene expression profile. This DNA sequence-independent inhibitory phenomenon suggests a common
mechanism unrelated to specific DNA sequence, and the
high frequency of unmethylated CpG dinucleotide
common to all bacterial DNA elements is a logical
candidate.
It has been well established that DNA carrying
multiple unmethylated CpG motifs is highly immunogenic,21 and has been used as an adjuvant for DNA
vaccines against cancers and other diseases.22 Unmethylated CpG motifs when complexed to cationic lipids
could elicit acute inflammatory reactions, resulting in the
activation of interferon-g (IFN-g) and tumor necrosis
factor-a (TNF-a), and necrosis-, apoptosis-mediated
DNA-transduced cell death, and consequently the loss
of transgene expression in vivo.8,9,23–25 Inflammatory
reactions stimulated by unmethylated CpG motifs are
an unwanted side effect for vectors used to express
therapeutic proteins. However, data from this and other
studies were not consistent with this as a major
mechanism of bacterial DNA inhibitory effect, especially
when vector DNA was delivered without cationic lipids.
In this and previous studies,1,16 we have repeatedly
demonstrated that 2fDNA always expressed a much
higher level of transgene products than standard
plasmids, although they shared exactly the same DNA
sequence, and the same amount of vector DNA persisted
in mouse liver long after administration and in no cases
was there a selective loss of cells harboring the bacterial
DNA. Thus, the preferential loss of vector DNA-containing cells is not a likely mechanism for the loss of
transgene expression. In addition, a cytotoxic response
was not observed after systemic delivery of either
oligonucleotides containing potent immunogenic unmethylated CpG motifs,9 or plasmid DNA8,26 alone.
Furthermore, a similar transient transgene expression
profile was observed in both immune-competent and
-deficient mice,2,27 excluding the possibility of the
involvement of a cytotoxic T-cell-mediated mechanism.
Finally, several other studies2,6,7 have provided evidence
suggesting that it is the promoter silencing, not the
vector DNA loss, that is responsible for the loss of
transgene expression.
An alternative noncytodestructive immunologic mechanism explaining the transgene silencing effect was
proposed by Qin et al,6 who demonstrated that IFN-g and
TNF-a could selectively inhibit several viral promoters
delivered in adenoviral, retroviral or plasmid vectors
in vitro and in vivo, while the constitutive cellular
promoter b-actin was less affected. It was also suggested
that the cytokines exert an inhibitory effect at multiple
levels including a direct effect at the transcriptional level
by turning off the viral promoter.10 It is well known that
cytokines, including IFN-g and TNF-a, are a part of the
innate immune system playing a key role in host
antiviral activities. It has been suggested that the
unmethylated CpG motifs can bind to a specific
Bacterial DNA in episomal transgene silencing
ZY Chen et al
intracellular receptor and activate the nuclear factor-kB
(NF-kB), which is a transcriptional factor responsible for
the induction of many cytokines.10,28 The role of CpG
dinucleotide-cytokine signaling in transgene silencing
appears to occur in other gene transfer studies where an
enhancement of transgene expression was demonstrated
by the administration of the anti-inflammatory agent
dexamethasone,9 and partial deletion of CpG motifs in
plasmid bacterial backbone.7 In the present study, all
three viral and four mammalian enhancer/promoters
examined were silenced in mouse liver when delivered
either as intact circular plasmid or as linearized DNA
generated by one cut through bacterial backbone,
whereas the enhanced transgene expression resulted
when dissociation between the two DNA elements
occurred, in either two fragment DNA, purified expression cassette, or minicircle group.16 Furthermore, our
results indicate that the silenced promoters (those that
connected with bacterial backbone) and the active
promoters (those that form DNA circle free of bacterial
DNA) could coexist in the same cell in two fragment
DNA group, indicating that a promoter remained active,
as long as it was not covalently connected with bacterial
backbone. Taken together, our result suggest that the
silencing is not the consequence of a direct inhibition, but
more likely resulted from an event which first occurred
at the bacterial DNA sequences, and then spread to the
downstream promoter, resulting in its inactivation, a
phenomenon as described by Kass et al,20 and discussed
earlier.
Another CpG motif-related mechanism was raised by
Hong et al,29 who provided preliminary evidence to
suggest that CpG dinucleotides in episomal vector DNA
could undergo de novo methylation in mammalian cells,
and suggested that this was a mechanism of episomal
transgene silencing. In eukaryotic cells, C5 methylation
of the CpG is one major mechanism of the regulation of
gene expression in chromosomes.30,31 Methylated CpG
dinucleotides can become the targets of a group of
cellular proteins, including CPM1 and CPM2, whose
binding would result in the condensation of DNA, and
consequently, the silencing of genes.32 If this was the
case, the problematic CpG clusters located within the
bacterial DNA sequence could trigger a nucleosome
condensation process,33,34 which could then spread to the
vicinity of the transgene, resulting in its silencing. It
remains possible that this in vivo CpG methylation is a
consequence of cytokine activation and a mechanism of
intracellular defense that evolved over time to silence
pathogenic organisms.
Materials and methods
Vector construction
Construction of the plasmids pRSV.hAAT.bpA (Figure 1)
expressing human AAT driven by RSV promoter, and
psApoE.hFIX þ Int A.bpA expressing the human FIX
under the control of a hybrid promoter/enhancer
composed of the human AAT promoter and the hepatic
locus control region (HCR) from the ApoE gene, were
described previously.1,35 Human CMV immediate-early
gene promoter derived from pcDNA3.1 (Invitrogen,
Carlsbad, CA, USA), human glyceraldehyde 30 -phosphate dehydrogenase promoter (GAPDH) and eukaryo-
tic initiation factor 4A1 promoter (heIF) from pDRIVEhGAPDH and pDRIVE-heIF4A1 (InvivoGene, San Diego,
CA, USA), respectively, were used to replace the RSV
promoter in pRSV.hAAT.bpA, resulting in plasmids
pCMV.hAAT.bpA, pGAPDH.hAAT.bpA and pheIF.hAAT.bpA. These promoters were amplified from
individual plasmids using PCR primers embedding a
HindIII or KpnI site, allowing them to replace the RSV
promoter in pRSV.hAAT.bpA. The hAAT.bpA fragment
was used to replace the EM7.Zeocin.SV40 poly-A
fragment in the pSV40/Zeo2 (Invitrogen, Carlsbad,
CA, USA), resulting in pMB1.SV40.hAAT. The
pCM.RSV.hAAT.bpA was constructed by replacing the
BspH1–BspH1 fragment encoding the Ampicillin resistance gene in pRSV.hAAT.bpA with the BspH1–BspH1
fragment encoding the CM resistance gene from
pBC.KS( þ ) (Stratagene, La Jolla, CA, USA). PCR
product of the DNA origin p15A using pACYC184
(New England Biolabs, Beverly, MA, USA) as template
was used to replace the ColE1 in pRSV.hAAT.bpA,
resulting in p15A.RSV.hAAT.bpA. The plasmid pEF1a.hFIX.hGHpA expressing the human FIX under the control
of the human elongation factor 1a promoter (EF1a), was
derived from pT.EF1a.hFIX36 by inserting a pair of DNA
oligonucleotides encoding an SpeI site into its NotI site
downstream of the human growth hormone polyadenylation signal (hGHpA), allowing the EF1a.hFIX.hGHpA
expression cassette to be released by SpeI digestion and
relocated into the SpeI site of pBS.KS.II (Stratagene). All
replacement of promoter/enhancers, antibiotic resistance
genes, the DNA origin or the expression cassette was
confirmed by DNA sequencing.
863
DNA preparation
All plasmid DNA were prepared using Qiagen (Valencia,
CA, USA) endotoxin-free kits. The two-fragment DNA
(2fDNA) form of pEF1a.hFIX.hGHpA (Figure 1) and
psApoE.hFIX þ Int A.bpA were prepared by digestion
of the plasmids with SpeI, which cut twice through
the bacterial DNA to release the expression cassettes. The
2fDNA from all other plasmids were prepared by
digestion of individual plasmid with XhoI. After gel
electrophoresis confirmation of completeness of the
digestion, the reactions were extracted once with
phenol:choroform:amyl alcohol (25:24:1), twice with
chloroform:amyl alcohol (24:1), and the DNA was
recovered by ethanol precipitation. To prepare the
purified expression cassette, the two DNA fragments
resulted from restriction digestion were separated by gel
electrophoresis, the desired expression cassettes were cut
out and the DNA was electroeluted from the gel and
recovered by ethanol precipitation. All DNAs were
dialyzed against TE (pH 8.0) for at least 24 h before
delivery to mice.
Animal studies
Each dose of DNA, ranging from 32 ng to 80 mg per
mouse, was dissolved in 1.8 ml of saline and infused into
mouse liver using the hydrodynamic protocol of Liu
et al13 and Zhan et al.37 C57L/6 mice (5–8 weeks old) from
Taconi Farms, Inc. (Germantown, NY, USA) were used.
Mice were bled periodically using a retro-orbital technique. Serum human FIX and human AAT were quantified
by ELISA as described previously.36
Gene Therapy
Bacterial DNA in episomal transgene silencing
ZY Chen et al
864
Southern blot analysis
Mouse liver DNA was prepared using a salt-out protocol.
In all, 20 mg of liver DNA from mice receiving different
forms of DNA derived from pRSV.hAAT.bpA was
digested with BglII, which did not cut the vector, or
HindIII, which cut once through the expression cassette,
resolved by gel electrophoresis, and probed with the
radio-labeled 1.35 kb hAAT cDNA fragment from
pRSV.hAAT.bpA (Figure 1). The radioactive vector
DNA bands were quantified by phosphoimager analysis
as described previously.1
Acknowledgements
We are extremely grateful to Theresa Storm for critical
reading of this paper. This work was supported by NIH
HL-64274.
References
1 Chen ZY et al. Linear DNAs concatemerize in vivo and result in
sustained transgene expression in mouse liver. Mol Ther 2001; 3:
403–410.
2 Gill DR et al. Increased persistence of lung gene expression using
plasmids containing the ubiquitin C or elongation factor 1alpha
promoter. Gene Therapy 2001; 8: 1539–1546.
3 Hartikka J et al. An improved plasmid DNA expression vector
for direct injection into skeletal muscle. Hum Gene Ther 1996; 7:
1205–1217.
4 Nicol F et al. Poly-L-glutamate, an anionic polymer, enhances
transgene expression for plasmids delivered by intramuscular
injection with in vivo electroporation. Gene Therapy 2002; 9:
1351–1358.
5 Wolff JA et al. Long-term persistence of plasmid DNA and
foreign gene expression in mouse muscle. Hum Mol Genet 1992;
1: 363–369.
6 Qin L et al. Promoter attenuation in gene therapy: interferongamma and tumor necrosis factor-alpha inhibit transgene
expression. Hum Gene Ther 1997; 8: 2019–2029.
7 Yew NS et al. High and sustained transgene expression in vivo
from plasmid vectors containing a hybrid ubiquitin promoter.
Mol Ther 2001; 4: 75–82.
8 Li S et al. Effect of immune response on gene transfer to the lung
via systemic administration of cationic lipidic vectors. Am J
Physiol 1999; 276: L796–804.
9 Tan Y, Li S, Pitt BR, Huang L. The inhibitory role of CpG
immunostimulatory motifs in cationic lipid vector-mediated
transgene expression in vivo. Hum Gene Ther 1999; 10: 2153–2161.
10 Paillard F. CpG: the double-edged sword. Hum Gene Ther 1999;
10: 2089–2090.
11 Miao CH, Thompson AR, Loeb K, Ye X. Long-term and
therapeutic-level hepatic gene expression of human factor IX
after naked plasmid transfer in vivo. Mol Ther 2001; 3: 947–957.
12 Yew NS et al. CpG-depleted plasmid DNA vectors with
enhanced safety and long-term gene expression in vivo. Mol
Ther 2002; 5: 731–738.
13 Liu F, Song Y, Liu D. Hydrodynamics-based transfection in
animals by systemic administration of plasmid DNA. Gene
Therapy 1999; 6: 1258–1266.
14 Kramer MG et al. In vitro and in vivo comparative study of
chimeric liver-specific promoters. Mol Ther 2003; 7: 375–385.
15 Critchlow SE, Jackson SP. DNA end-joining: from yeast to man.
Trends Biochem Sci 1998; 23: 394–398.
Gene Therapy
16 Chen ZY, He CY, Ehrhrdt A, Kay MA. Minicircle DNA vectors
devoid of bacterial DNA result in persistent and high level
transgene expression in vivo. Mol Ther 2003; 8: 495–500.
17 Wolffe AP. Histone modification and transcriptional
competence. In: Driel RV, Otte AP (eds) Nuclear Organization,
Chromatin Structure, and Gene Expression. Oxford University
Press: Oxford, New York, Tokyo, 1997, pp 40–57.
18 Utley RT et al. Transcriptional activators direct histone
acetyltransferase complexes to nucleosomes. Nature 1998; 394:
498–502.
19 Noma K, Allis CD, Grewal SI. Transitions in distinct histone H3
methylation patterns at the heterochromatin domain
boundaries. Science 2001; 293: 1150–1155.
20 Kass SU, Landsberger N, Wolffe AP. DNA methylation directs a
time-dependent repression of transcription initiation. Curr Biol
1997; 7: 157–165.
21 Krieg AM et al. CpG motifs in bacterial DNA trigger direct B-cell
activation. Nature 1995; 374: 546–549.
22 Krieg AM et al. Sequence motifs in adenoviral DNA block
immune activation by stimulatory CpG motifs. Proc Natl Acad Sci
USA 1998; 95: 12631–12636.
23 Schwartz DA et al. CpG motifs in bacterial DNA cause
inflammation in the lower respiratory tract. J Clin Invest 1997;
100: 68–73.
24 Yew NS et al. Contribution of plasmid DNA to inflammation in
the lung after administration of cationic lipid:pDNA complexes.
Hum Gene Ther 1999; 10: 223–234.
25 Freimark BD et al. Cationic lipids enhance cytokine and
cell influx levels in the lung following administration of
plasmid: cationic lipid complexes. J Immunol 1998; 160:
4580–4586.
26 Scheule RK et al. Basis of pulmonary toxicity associated with
cationic lipid-mediated gene transfer to the mammalian lung.
Hum Gene Ther 1997; 8: 689–707.
27 Huang CY et al. Enhancements in gene expression by the choice
of plasmid DNA formulations containing neutral polymeric
excipients. J Pharmacol Sci 2002; 91: 1371–1381.
28 Hemmi H et al. A Toll-like receptor recognizes bacterial DNA.
Nature 2000; 408: 740–745.
29 Hong K, Sherley J, Lauffenburger DA. Methylation of episomal
plasmids as a barrier to transient gene expression via a synthetic
delivery vector. Biomol Eng 2001; 18: 185–192.
30 Holliday R, Pugh JE. DNA modification mechanisms and gene
activity during development. Science 1975; 187: 226–232.
31 Sager R, Kitchin R. Selective silencing of eukaryotic DNA.
Science 1975; 189: 426–433.
32 Bird AP, Wolffe AP. Methylation-induced repression – belts,
braces, and chromatin. Cell 1999; 99: 451–454.
33 Schubeler D et al. Genomic targeting of methylated DNA:
influence of methylation on transcription, replication, chromatin
structure, and histone acetylation. Mol Cell Biol 2000; 20:
9103–9112.
34 Mummaneni P, Walker KA, Bishop PL, Turker MS. Epigenetic
gene inactivation induced by a cis-acting methylation center.
J Biol Chem 1995; 270: 788–792.
35 Miao CH et al. Inclusion of the hepatic locus control region, an
intron, and untranslated region increases and stabilizes hepatic
factor IX gene expression in vivo but not in vitro. Mol Ther 2000; 1:
522–532.
36 Yant SR et al. Somatic integration and long-term transgene
expression in normal and haemophilic mice using a DNA
transposon system. Nat Genet 2000; 25: 35–41.
37 Zhang G, Budker V, Wolff JA. High levels of foreign gene
expression in hepatocytes after tail vein injections of naked
plasmid DNA. Hum Gene Ther 1999; 10: 1735–1737.